Abstract:• A data augmentation method is proposed to train deep neural nets with limited data. • Random image patching is used to generate new images for training. • Edge matching cost is introduced to select optimum generated images. • Mean accuracy increases from 91.01, 97.99 to 94.02, 98.51 in two datasets. • Mean IOU increases from 63.59, 74.26 to 70.77, 77.09 in two datasets. Deep learning methods such as convolutional neural networks (CNN) have become popular for addressing crops and weeds classification problems in agricultural robotics. However, to have satisfactory performance and avoid overfitting, training deep neural nets typically requires thousands of labeled images. This leads to tedious pixelwise labeling for semantic segmentation. In this paper, we hinge on the recent development in data augmentation and utilize the concept further for semantic segmentation and classification of crops and weeds. To be specific, we propose a novel data augmentation framework, based on the random image cropping and patching (RICAP) method, which is originally designed to augment data for generic image classification. The proposed framework introduces novel enhancements to the original RICAP so that it can be effectively used for data augmentation of semantic segmentation tasks. We evaluate the proposed methodology on two datasets from different farms. Comprehensive experimental evaluations and ablation studies show that the proposed framework can effectively improve segmentation accuracies, and the enhancements made over the original RICAP actually contribute to the performance gain. On average, the proposed method increases the mean accuracy and mean intersection over union (IOU) of the deep neural net with the conventional data augmentation (random flipping, rotation and colour jitter) from 91.01 to 94.02 and from 63.59 to 70.77 respectively for Narrabri dataset, and from 97.99 to 98.51 and from 74.26 to 77.09 respectively for Bonn dataset. The limitation of the proposed method, especially when a large number of training data is available, has also been discussed.

Data Augmentation for Deep Learning Based Semantic Segmentation and Crop-Weed Classification in Agricultural Robotics

A Deep Learning Image Augmentation Method for Field Agriculture

Optimizing Data Augmentation for Semantic Segmentation on Small-Scale Dataset

Data Augmentation for Deep Learning based Cattle Segmentation in Precision Livestock Farming

Image-to-Image Translation-Based Data Augmentation for Improving Crop/Weed Classification Models for Precision Agriculture Applications

Shape and Style GAN-based Multispectral Data Augmentation for Crop/Weed Segmentation in Precision Farming

Fine-tuning Convolutional Neural Network with Transfer Learning for Semantic Segmentation of Ground-Level Oilseed Rape Images in a Field with High Weed Pressure.

Bagging Improves the Performance of Deep Learning-Based Semantic Segmentation with Limited Labeled Images: A Case Study of Crop Segmentation for High-Throughput Plant Phenotyping

A Modified U-Net with a Specific Data Argumentation Method for Semantic Segmentation of Weed Images in the Field.

A Data-Related Patch Proposal for Semantic Segmentation of Aerial Images

Towards agricultural autonomy: crop row detection under varying field conditions using deep learning

Semantic Segmentation of Crop and Weed using an Encoder-Decoder Network and Image Enhancement Method under Uncontrolled Outdoor Illumination

Transferring learned patterns from ground-based field imagery to predict UAV-based imagery for crop and weed semantic segmentation in precision crop farming

Image patch-based deep learning approach for crop and weed recognition

Multi-Spectral Image Synthesis for Crop/Weed Segmentation in Precision Farming

Towards Domain Generalization in Crop and Weed Segmentation for Precision Farming Robots

Data Augmentation in Classification and Segmentation: A Survey and New Strategies

Deep Data Augmentation for Weed Recognition Enhancement: A Diffusion Probabilistic Model and Transfer Learning Based Approach

3D Point Cloud Semantic Segmentation Toward Large-Scale Unstructured Agricultural Scene Classification

3D data-augmentation methods for semantic segmentation of tomato plant parts

Crop and Weed Segmentation and Fractal Dimension Estimation Using Small Training Data in Heterogeneous Data Environment